Mp3 export much faster with larger Navigator
Hello,
I have observed that exporting my scores to mp3 can be very slow when I have the Navigator window relatively small (low). But if I make the Navigator larger, 1.5x or 2x as high, exporting to mp3 will be much faster!
I discovered this by accident, and it strikes me.
Is there an explanation for it?
I would rather not have to enlarge the Navigator window for exporting to mp3, as it will prevent me from seeing all the staves in my score.
| Attachment | Size |
|---|---|
| NavigatorSmaller.jpg | 83.23 KB |
| NavigatorLarger.jpg | 85.85 KB |
Comments
The current version is 2.3.2 if you don't have it. In previous versions, when the navigator was at the default size, MuseScore slowed down considerably, making it a bit smaller fixed this. This was fixed at some time before the current version. I'm curious if you're using 2.3.2. If you are, then someone needs to look at this and make sure the problem does not exist in version 3, which about to start beta testing.
In reply to The current version is 2.3.2… by mike320
Yes, I am using 2.3.2.
It's interesting to hear that making the navigator smaller made the mp3 export faster, as well.
Just tried it, and it does.
How come that the navigator window size has any impact on the mp3 export speed?
In reply to Yes, I am using 2.3.2. It's… by pseller
This is the first I've heard of the navigator affecting mp3 export, the navigator in previous versions slowed down note input and updating the score. Can you test this on a version 3 nightly build located here. Choose the second one, the first one is old and for converting an incompatible version. Note, that at the last report, there is a bug that will crash the program if several mp3's are exported, so I suggest that you test no more than 2 mp3 exports before closing and reopening the nightly. Also, there is no installation necessary, unzip the files and navigate to the /bin/ folder and run the nightly.exe file.
In reply to This is the first I've heard… by mike320
I tried it out with various sizes of the navigator window. It seems to make no difference in MS3.
And I'm not getting differences in MS2.3.2 any longer either - at least not with the score that I had already opened and changed the navigator window size.
I'm still wondering how it ever was able to make a difference on the mp3 export speed.
That is the only export that takes significant time, i.e. around 3 minutes for my 96 measure, 7 instrument tango score of Re Fa Si.
I'm uploading that score, in case you want to take measurements.
In reply to I tried it out with various… by pseller
BTW, maybe it is important to mention that I am using various soundfonts (see attachment).
I included those before exporting Re Fa Si using MS3, just as they are included in my MS2.3.2.
In reply to BTW, maybe it is important… by pseller
I'm curious if the anomaly has to do with this:
You opened MuseScore with the Navigator large and quickly exported to mp3, then shrunk the navigator and exported to mp3 again. If I'm correct, it seemed the Navigator was the culprit but in truth the culprit would have been the MuseScore General.sf3. It takes a bit of time to unpack an sf3 sound font, and this is what actually slowed down the original export. Now, you've given the sf3 time to unpack and no longer see the delay. The reason you didn't see the delay in version 3 is that you had to find the navigator to open it and the sf3 had time to unpack.
That's my theory.
In reply to I'm curious if the anomaly… by mike320
That's an interesting theory.
However, I think I have also got those delays after having worked on a score for hours.
And resizing the Navigator window made a difference of something like factor 5-10. So, when I notice that even the first 1-2% of exporting to mp3 take more than 10 seconds, I resize, and then i will get the first 10% in less than 5 seconds.
In any case, I didn't get very different export to mp3 times with MS3.
Does a sf3 need to be unpacked every time you want to export to mp3?
Is it different with sf2?
Mostly or actually exclusively, I am relying on the top two soundfonts in the list I sent the screenshot of.
All my instruments use soundfonts out of those two collections.
So, I might as well get rid of the sf3, if that might make a difference ...
In reply to That's an interesting theory… by pseller
The sf3 is unpacked when it's loaded. If it's a default, that's every time you open MuseScore. If you add it after a while, it get unpacked when you add it. sf2's aren't compressed so they are ready as soon as they're loaded.
Of course it could always be another process on you computer slowing down the export occassionally.
In reply to The sf3 is unpacked when it… by mike320
I did see a strong correlation between changing the navigator window size and the export speed for mp3s. It appears very unlikely that it's not to be due to the change of the navigator window size.
Nothing much else is happening on my computer, at least not that I could see changing substantially during that short period of time.
I just tried out deleting the sf3, and it appears that MS3 managed to export the mp3 about 20-30% faster.
It is conspicuous though that the progress in terms of percentage exported is faster during the first 50 % than during the last 50 %. That probably indicates that those percentages are estimates rather than measurements, right?
Is there any other test which would give us more clues?
In reply to I did see a strong… by pseller
There are others more knowledgeable about mp3 export than me, so almost I'm done with my knowledge about the difference between export speeds. MuseScore 3 is designed to do things faster than version 2, so it's not surprising the MS3 export is faster.
You could try one other thing. Export the mp3 two times in a row without changing anything else and see if there is a speed difference. Even keep the sf3 loaded and give it some time to unpack when you first open MS. Don't do this in version 3 because there is a memory leak in mp3 export that would cause the second export to be slower. Enough exports will cause MS3 to crash.
In reply to There are others more… by mike320
What memory leak are you talking about?
In reply to What memory leak are you… by Jojo-Schmitz
I had #276356: Memory leak and crash after multiple audio exports (MP3, WAV, OGG, FLAC) in mind. I didn't realize it had been fixed.
In reply to I had #276356: Memory leak… by mike320
Phew, I feared there was another one ;-)
In reply to I did see a strong… by pseller
It is conspicuous though that the progress in terms of percentage exported is faster during the first 50 % than during the last 50 %. That probably indicates that those percentages are estimates rather than measurements, right?
Half true :)
Exporting to mp3 is a two step process and each step is given 50% of the total bar.
The first step is the actual rendering of the score using the soundfont samples; this is thus influenced by which soundfonts you have loaded.
The second step is peak normalization in which the result of the previous step is analysed and then "scaled" up so to maximize the volume of the loudest point within that score. Here soundfonts have no influence at all anymore.